The acquisition of a conditioned fear response is adaptive, as it enables the organism to appropriately respond to predictors of aversive events. Consequently, the absence of predictive cues can be used as a signal for safety. We aimed to study whether deficient fear conditioning might lead to maladaptive fear. Following previous work, we predicted that failure to learn the CS-US association would result in higher contextual fear, and that participants who failed to learn would tend to exhibit higher trait anxiety. Conditioning took place in a virtual environment with two contexts. In one of the two contexts, offset of a CS (light) was associated with a shock. Each participant visited two places (a house and an apartment) in each of 12 blocks. In one of these places shocks were administered at the offset of an 8-s period of lights on (CS). The results showed that half of the participants demonstrated differential shock expectancy between situations in the shock context in which the CS was present versus absent. This indicates that these participants learned the contingencies between the shocks and both the context and the light CS. In contrast, the other half of the participants learned only the association with the context. As predicted, learning the CS-US contingency resulted in reduced self-reported fear in the absence of the CS in the danger context compared to the presence of the CS. On the other hand, participants who failed to learn the association displayed a sustained aversive state throughout the duration of the danger context. Skin conductance measures confirmed this pattern of results. Fear-potentiated startle during the threat context compared to the safe context was significant in both groups, while startle was only potentiated during the CS in the threat context in the group that learned the CS association (trend-level significant). Finally, scores on Spielberger's self-report scale of trait anxiety tended to be higher in the group of participants who did not learn the CS-shock association in the danger context compared to participants who did. In conclusion, these results confirm higher contextual fear in participants who did not acquire a conditioned response to the cue in comparison to participants who did. Also, virtual reality contexts provide a useful tool for the study of context conditioning.